I learned about it at university. The idea is to measure the difference between periods of the short and the long pendulums. In particular, how many swings it takes for their swings to match (both pendulums arriving at the apex at the same time).
oh, and if it takes a different amount of time, something interfered, I'll have to try and make a simulation using the formulas on that page, then i can see if it can be copied with 6|9DOF, MEMS vibrate and i think you can control how fast, or at least track how fast over a period of time. i think you'd need 3, but i haven't messed with the expensive bosch sensors in a while.
Not even that, the period of a pendulum is 2pisqrt(L/g), so the interference will likely affect _both_ pendulums similarly.
The cool thing is that two pendulums can be modeled as simple sine waves (and you can do Fourier series to account for non-linearity of real pendulums), so the sum of them is _also_ a sine wave. And if you get its period, you can easily compute the `g` (since the ratio of lengths of pendulums is known to a good precision).
Detecting the period is also easy, you can have two electric contacts at the apex of the swing, and the circuit gets completed when both pendulums arrive at the same exact time to touch both contacts.
